Robert J. Colvin

Roger C. Su, Robert J. Colvin

Memory consistency models define the order in which accesses to shared memory in a concurrent system may be observed to occur. Such models are a necessity since program order is not a reliable indicator of execution order, due to microarchitectural features or compiler transformations. Concurrent programming, already a challenging task, is thus made even harder when weak memory effects must be addressed. A rigorous specification of weak memory models is therefore essential to make this problem tractable for developers of safety- and security-critical, low-level software. In this paper we survey the field of formalisations of weak memory models, including their specification, their effects on execution, and tools and inference systems for reasoning about code. To assist the discussion we also provide an introduction to two styles of formal representation found commonly in the literature (using a much simplified version of Intel's x86 as the example): a step-by-step construction of traces of the system (operational semantics); and with respect to relations between memory events (axiomatic semantics). The survey covers some long-standing hardware features that lead to observable weak behaviours, a description of historical developments in practice and in theory, an overview of computability and complexity results, and outlines current and future directions in the field.

Robert J. Colvin, Kirsten Winter

Reasoning about correctness and security of software is increasingly difficult due to the complexity of modern microarchitectural features such as out-of-order execution. A class of security vulnerabilities termed Spectre that exploits side effects of speculative, out-of-order execution was announced in 2018 and has since drawn much attention. In this paper we formalise speculative execution and its side effects with the intention of allowing speculation to be reasoned about abstractly at the program level, limiting the exposure to processor-specific or low-level semantics. To this end we encode and expose speculative execution explicitly in the programming language, rather than solely in the operational semantics; as a result the effects of speculative execution are captured by redefining the meaning of a conditional statement, and introducing novel language constructs that model transient execution of an alternative branch. We add an abstract cache to the global state of the system, and derive some general refinement rules that expose cache side effects due to speculative loads. Underlying this extension is a semantic model that is based on instruction-level parallelism. The rules are encoded in a simulation tool, which we use to analyse an abstract specification of a Spectre attack and vulnerable code fragments.

Ian J. Hayes, Larissa A. Meinicke, Kirsten Winter et al.

This research started with an algebra for reasoning about rely/guarantee concurrency for a shared memory model. The approach taken led to a more abstract algebra of atomic steps, in which atomic steps synchronise (rather than interleave) when composed in parallel. The algebra of rely/guarantee concurrency then becomes an instantiation of the more abstract algebra. Many of the core properties needed for rely/guarantee reasoning can be shown to hold in the abstract algebra where their proofs are simpler and hence allow a higher degree of automation. The algebra has been encoded in Isabelle/HOL to provide a basis for tool support for program verification. In rely/guarantee concurrency, programs are specified to guarantee certain behaviours until assumptions about the behaviour of their environment are violated. When assumptions are violated, program behaviour is unconstrained (aborting), and guarantees need no longer hold. To support these guarantees a second synchronous operator, weak conjunction, was introduced: both processes in a weak conjunction must agree to take each atomic step, unless one aborts in which case the whole aborts. In developing the laws for parallel and weak conjunction we found many properties were shared by the operators and that the proofs of many laws were essentially the same. This insight led to the idea of generalising synchronisation to an abstract operator with only the axioms that are shared by the parallel and weak conjunction operator, so that those two operators can be viewed as instantiations of the abstract synchronisation operator. The main differences between parallel and weak conjunction are how they combine individual atomic steps; that is left open in the axioms for the abstract operator.